Harnessing machine learning for enhanced thermal insulation and energy efficiency in buildings worldwide

Despite numerous studies on building energy efficiency, accurately predicting energy consumption remains challenging due to the complex interaction between climate variability and the diversity of thermal insulators. Many existing models do not simultaneously address both energy consumption and thermal comfort across different climates, limiting their applicability for holistic building performance optimization. This research addresses these gaps by developing a machine learning-based model capable of predicting two key indices: energy consumption (Φ) and thermal comfort (f) across various climates, defined by the Köppen-Geiger classification. The goal is to identify the optimal thermal insulator for each climate, contributing to a more adaptable and climate-responsive approach to energy efficiency. Four machine learning (ML) algorithms were evaluated using data generated from numerical simulations in MATLAB: Support Vector Machine (SVM), Artificial Neural Network (ANN), Random Forest (RF), and Extreme Gradient Boosting (XGBoost). Results demonstrate that the XGBoost and RF models deliver superior performance, with accuracies of 96.53 % and 94.6 % for Φ and 97.82 % and 97.5 % for f, respectively. In contrast, SVM showed lower predictive power, achieving 60.56 % for Φ and 87.14 % for f. These findings underscore the effectiveness of RF and XGBoost in accurately predicting both energy consumption and thermal comfort. The proposed model offers a significant contribution by providing a reliable tool for optimizing building design strategies across diverse climatic conditions, ultimately promoting energy-efficient and comfortable living environments.